mlcommons / training_results_v0.5

This repository contains the results and code for the MLPerf™ Training v0.5 benchmark.
https://mlcommons.org/en/training-normal-05/
Apache License 2.0
35 stars 54 forks source link

Nvidia/translation ModuleNotFoundError: No module named 'strided_batched_gemm' #4

Closed dujiangsu closed 5 years ago

dujiangsu commented 5 years ago

image

The directory fairseq has some cpp or cu files which cannot be imported.

dujiangsu commented 5 years ago

Also, the fairseq used in your project is totally different from the one I git clone from pytorch/fairseq. Could you please tell me which version you use? Thanks